Thanks for taking the time to do these tests David!

I've conducted the following test: creating a perspective cam that matches the orthographic cam almost 1:1 by playing with focal length etc. The result is: the perspective cam renders completely differently from the ortho cam even through they're framing exactly the same thing.

Therefore, I think it disproves this issue being created by the lights. What do you think ?

I guess I solved my problem by replicating the ortho cam's results with a perspective cam but I'm still curious why would the two cameras render completely different results...